As You Like It: Analyzing Public Sentiment with Twitter Engagement Metadata

Abstract

As the role of technology within our society continues to grow, social media becomes an increasingly important factor in the way information is distributed and consumed. In the past decade there have been many studies that explore how public sentiment from social media can provide greater insights for a plethora of applications, ranging from identifying product issues to trading equities.

This thesis aims to better understand if and how sentiment models can be enriched through the incorporation of engagement metadata available on Twitter. Two novel public sentiment models, weighted by engagement (i.e. favorites and retweets) are proposed and examined. By analyzing thousands of publicly available tweets pertaining to four consumer companies, we conduct a comparison study to understand the differences between our Engagement Weighted Sentiment Models versus a traditional text only model. We explore how these findings have applications to decision making within consumer markets for different entities.